Tech companies that heavily relied on internal AI are now reevaluating due to the escalating costs associated with intensive AI usage. Uber, for instance, revealed that it had already depleted its entire 2026 AI budget within the first four months of the year, prompting concerns about justifying these internal AI expenses. OpenAI’s CEO, Sam Altman, also highlighted the significant financial burden AI costs pose for their customers.
This issue extends beyond major corporations, as Canadian startup leaders discussed the challenges of managing increasing internal AI expenses at a recent conference. The focus has now shifted towards implementing better cost-tracking mechanisms and utilizing AI in a more strategic manner. However, this shift in spending raises questions about the impact on the sky-high valuations of AI companies.
The surge in expenses primarily stems from the use of “tokens,” which represent the units of data needed to input prompts for AI and obtain outputs. Companies have been consuming substantial amounts of tokens, driven by the trend of “tokenmaxxing,” which essentially reflects the cost of user interactions with AI.
While the overall cost of real-world AI applications, known as inferences, has decreased, tech firms are increasingly employing AI for intricate, “agentic” tasks such as coding and complex reasoning processes. This contrasts with simpler interactions like seeking recipe advice from ChatGPT, as these advanced tasks require significantly more tokens.
Previously, many companies encouraged employees to experiment extensively with AI, leading to practices like “tokenmaxxing” to demonstrate productivity. However, faced with the financial implications of heavy token use, some businesses are now revising their expenditure strategies. For example, Uber recently imposed a monthly cap of $1,500 per employee per coding tool.
To navigate the balance between innovation and cost control, businesses are now delving into AI “tokenomics,” which involves a deeper understanding of token costs and strategic utilization of AI resources for predictable financial outcomes. Nestor Maslej, a consulting firm CEO and former editor-in-chief of Stanford University’s AI Index Report, recommends conducting micro-sized experiments to identify AI’s practical utility and comparing its efficiency and cost-effectiveness against human counterparts.
Ultimately, the AI sector faces a pivotal moment as companies weigh the returns on investment against the expenses incurred for complex AI applications. The evolving landscape may force companies to find a delicate balance between recouping token costs and maintaining market share in a fiercely competitive environment. Strategies such as adjusting pricing models based on token usage, as seen with various industry players, indicate the ongoing evolution and price adjustments within the AI market. Maslej suggests that despite the challenges, there remains a level of cost that businesses are willing to bear for the benefits of AI technology.
